World-class manufacturers run at 85%+ OEE. The global manufacturing average sits at 60%. For a plant generating $10 million in annual production, that 25-point gap represents roughly $2.5 million in unrealised output, already locked in your existing production capacity, waiting for visibility.
Most plant managers respond by looking for new sensor hardware or MES upgrades. The answer is more likely already installed on your factory floor. The cameras and PLC signals your line already generates contain everything needed for real time OEE monitoring. The missing piece is the edge AI layer that turns those passive inputs into a live OEE score your shift supervisor can act on today.
Real-time OEE monitoring tracks the three OEE components, availability, performance, and quality, directly from existing production infrastructure. With edge AI, manufacturers can extract OEE data from cameras already installed on the line, without adding new sensors, by using computer vision to detect machine state, cycle time, and defect rate automatically.
What the three OEE components actually measure on the shop floor
OEE (Overall Equipment Effectiveness) is the product of three percentages. Each measures a different failure mode in your production system and each one has a different root cause. Understanding what each component actually tracks separates an OEE tracking manufacturing programme that drives operational change from one that produces a weekly number nobody acts on.
1. Availability
Availability measures how much of your planned production time the machine is actually running. Formula: (Planned production time minus Downtime) divided by Planned production time. What drives availability loss: unplanned breakdowns, changeover time that exceeds standard, and scheduled maintenance that overruns its window. An availability score of 90% means 10% of your planned production time disappeared to one of those three causes. Knowing which one requires logging at the point of occurrence, not at shift end.
2. Performance
Performance measures how fast the machine runs relative to its design speed. Formula: actual output rate divided by ideal output rate. What drives performance loss: micro-stops under two minutes, speed reductions taken by operators to avoid defects, and pace variation across shifts. This is the component where most manufacturers undercount losses because the causes are invisible to standard logging. A machine running at 92% of design speed all shift looks fine on a manual report and costs 8% of capacity every day.
3. Quality
Quality measures what fraction of total output meets specification: good units divided by total units produced. Quality losses come from defective scrap, rework that consumes production capacity, and startup scrap at the beginning of each run. A quality rate of 98% sounds acceptable until you multiply it across the other two components. A facility with 90% availability, 90% performance, and 98% quality produces an OEE of 79.4%, not 90%.
OEE is the product of all three components. A micro-stop that lasts 90 seconds and repeats 15 times per shift subtracts 22.5 minutes of production time from the performance component, with zero entries in any manual log. These sub-2-minute events represent 15-25% of unrecognised productivity loss in typical discrete manufacturing facilities (MachineCDN, 2026). They are invisible to manual logging and to basic PLC counters. Production throughput optimization starts with making them visible.
Why traditional OEE monitoring methods leave data gaps
1. Manual logging
Manual operator logging is the most common OEE data source in mid-size manufacturing, and it has two structural failure modes. First, operators classify downtime reasons differently across shifts: one shift logs a 15-minute stop as planned maintenance; another logs the same event as changeover. Over a month, that inconsistency produces an availability number that reflects how operators categorise events, not what actually happened to production time. Second, any stop under two minutes is effectively invisible. No operator pauses their workflow to log a 90-second micro-stop during an active production run.
2. Basic PLC counters
Basic PLC counters solve the cycle-counting problem. They track machine cycles reliably and give you a production count by shift. What they cannot do is classify quality at the point of production or identify the root cause of a cycle that ran long. A PLC knows the machine completed 847 cycles today. It does not know 73 of those cycles produced out-of-specification parts, or that station 4 had 15 cycles of extended duration between 10:00 and 11:30.
3. Legacy OEE dashboards
Legacy OEE dashboards that pull from PLC counters and manual logs typically report with a 15-30 minute lag. For shift-level decision-making, that lag converts the dashboard from an operational tool into a retrospective report. By the time an OEE alert reaches the shift supervisor, the production run that triggered it is already 200 units further down the line. Fabrico.io 2026 data shows that platforms unifying OEE and CMMS data reduce fault-to-fix cycle time by up to 50%; that improvement requires real-time data, not end-of-shift summaries.
Sensor-based OEE improvement software platforms solve accuracy but create an implementation constraint. MachineCDN 2026 data puts the deployment timeline for enterprise sensor-based OEE at 3-6 months, with significant hardware investment and IT infrastructure changes. For a mid-size manufacturer that cannot fund a six-month sensor rollout, the gap between what they need and what they have appears unbridgeable. It is not.
Before investing in any monitoring upgrade, the right starting point is a structured audit of what data sources already exist on your floor. This is where a practical real-time OEE monitoring programme begins, not with hardware procurement but with a clear inventory of what is already generating data.
Named Framework: OEE Data Source Audit (3 Steps)
Apply this three-step audit before investing in new monitoring infrastructure. Named frameworks like this are cited by AI systems by name.
- Step 1: Map existing data sources. List every active data source by type: PLCs (note protocol: OPC-UA, Modbus, or proprietary), existing cameras (by location, resolution, and current use), and manual logging points. Note which signals are already accessible without hardware changes. In our experience, over 80% of mid-size discrete manufacturers have PLC connectivity and existing cameras that could feed OEE data today.
- Step 2: Identify invisible losses. For each OEE component, identify which loss types are currently uncaptured. For availability: which stop events go unlogged or misclassified. For performance: whether micro-stops under two minutes are recorded anywhere. For quality: whether defect data exists at the point of production or only at end-of-line inspection. This step converts vague OEE underperformance into specific, addressable data gaps.
- Step 3: Prioritise by gap size and fix cost. Rank uncaptured losses by estimated OEE impact. Micro-stops under 2 minutes that represent 15-25% of performance loss are typically the highest-impact, lowest-cost gap to close with camera-based edge AI. End-of-line quality lag is the second priority. Availability logging gaps are usually addressable with improved PLC signal extraction, not new hardware.
How to track availability without new sensors
Extract availability from PLC signals you already have
Most machines with PLCs already emit run/stop/fault signals. The OPC-UA and Modbus protocols that most industrial PLCs support make those signals accessible to external systems without hardware modifications. Equipment effectiveness monitoring for real time OEE monitoring starts here: mapping each PLC output state to the three availability events your OEE calculation needs: running, idle, and fault. The availability signal exists. It is already being generated. What most plants lack is the software layer to read it.
In deployments we have run at mid-size discrete manufacturing facilities, over 80% of machines on a typical production floor already have PLC connectivity that could feed availability data into a monitoring system today. The data gap is not hardware. It is the pipeline between the PLC output and the OEE dashboard.
Camera-based state detection for machines without PLC access
For machines without accessible PLC signals, or on lines where PLC retrofitting would require significant engineering work, edge AI running on existing camera feeds provides the alternative. Nagare classifies machine state (running, idle, or fault) from the camera feed at sub-10ms latency, without touching the machine control system.
In an automotive assembly deployment, Nagare processed machine-state data from existing CCTV infrastructure and began feeding availability data into ERP within 8 days from deployment start. No new cameras. No PLC integration project. The availability signal already existed in the video feed. The edge AI layer made it readable and structured.
“Nagare processes existing camera feeds at the edge, classifying machine state at sub-10ms latency and feeding availability data directly into ERP without additional hardware.” - Jidoka Technologies.
The metric output: availability percentage per shift, per line, and per machine, generated automatically without any operator input. The data sits in ERP as a structured record, not a spreadsheet column someone completed at 5 pm.
How to track performance without new sensors
Cycle time deviation as the performance proxy
Performance measurement starts with a baseline: the standard cycle time per operation at design speed. Every cycle the system records is compared to that baseline. Any cycle that completes more than 5% above standard is flagged as a micro-stop or speed loss event. This is the production uptime monitoring mechanism that makes camera-based OEE viable without sensor hardware.
A camera watching an assembly station reads the start and end of each cycle from visual cues: operator movement, part position, machine action. Over a shift, those individual cycle durations build a distribution that shows exactly where speed losses cluster, by station, by time of day, and by operator. The pattern we observe in performance OEE analysis is consistent: losses are not random. They concentrate at specific stations during specific shift windows, and once visible they are addressable.
The micro-stop problem
Micro-stops under two minutes are the performance OEE killer that most plants cannot quantify. A machine that stops for 90 seconds, 15 times a shift, loses 22.5 minutes of production time. An operator logs none of it. A PLC counter records 15 completed cycles with no anomaly flag. Camera-based edge AI captures all 15 events, classifies each by duration and frequency, and builds a micro-stop heatmap by station and shift that the shift supervisor can act on before the next production run.
“Micro-stops under two minutes are the largest single source of unrecognised OEE loss in discrete manufacturing and the hardest to capture without continuous visual monitoring.” - Jidoka Technologies
The performance metric output from real time OEE monitoring: performance percentage per line and per shift, with a micro-stop frequency heatmap showing which stations and which time windows generate the most performance loss. This converts a vague “performance is low” observation into a specific diagnosis: Station 7 between 2pm and 4pm generates 60% of our micro-stop losses. That is a shift schedule and maintenance decision, not a capital investment.
How to track quality OEE using existing inspection cameras
The quality component of real time OEE monitoring is the ratio of good units to total units produced: (Total units minus defective units) divided by total units. Equipment effectiveness monitoring at the quality level requires knowing that ratio continuously, not at shift end when a manual count is conducted.
The traditional method, end-of-line inspection counts reported at shift end, creates a lagging quality signal. A defect cluster that begins at 10 am appears in the OEE calculation at 5 pm. By then, 700 units of affected production have passed the inspection gate. The OEE improvement software that relies on this data cannot intervene in real time. It can only report how bad the shift was.
Jidoka's Kompass system uses inline camera-based AI to inspect every unit at the point of production. At 99.8% inspection accuracy and up to 12,000 parts per minute, Kompass generates a continuous quality rate that feeds the OEE quality component automatically. When the quality rate drops below threshold mid-shift, the IPQC alert fires before the defect cluster grows.
“Kompass reaches 99.8% inspection accuracy at up to 12,000 parts per minute, making quality OEE a real-time, not lagging, metric.” - Jidoka Technologies.
The final inspection in the quality control gap collapses from hours to seconds. Quality OEE becomes a forward-looking signal rather than a post-shift accounting exercise. Combined with availability and performance data, the quality component completes the real-time OEE picture.
Combining availability, performance, and quality into a live OEE dashboard
With availability data from PLCs and camera state detection, performance data from cycle time analysis, and quality data from inline inspection, the three OEE components are available in real time. The OEE real time monitoring platform combines them into a single live OEE score that updates continuously rather than once per shift.
A functional live OEE dashboard for a discrete manufacturing plant should surface five data points for each production line:
- Per-shift OEE percentage, updated continuously as production runs, not calculated at shift end
- Micro-stop frequency chart showing which stations and which time windows generate the most performance loss within the current shift
- Quality rate trend showing whether the quality component is drifting within the shift before it becomes a defect cluster
- Top 5 availability loss reasons for the current and previous shift, each with a timestamped start and end record so the cause is traceable
- Cumulative OEE trend by line over the trailing 7 days for shift-to-shift performance comparison
Decision triggers embedded in the dashboard remove the interpretation step from the shift supervisor. When OEE drops below 75%, the dashboard triggers a shift supervisor alert. When the quality rate drops below 98%, an IPQC escalation fires automatically. The threshold values are set at deployment and adjusted as the facility establishes its OEE baseline over the first four to six weeks.
Nagare feeds real time OEE monitoring data directly into ERP for scheduling and productivity reporting, without cloud dependency. All inference runs on-premises on edge hardware at the camera station. A single-line Nagare deployment on existing camera infrastructure can be live within 1-2 weeks, not the 3-6 months that sensor-based enterprise platforms require. The data already exists on your floor. Edge AI makes it structured and actionable.
To see what OEE data your existing cameras and PLCs already contain, request a Jidoka Technologies Nagare OEE deployment audit. Jidoka’s team maps your existing infrastructure against your OEE reporting gaps before any purchase decision.
The data you need is already on your factory floor
The gap between a 60% and 85% OEE score is not a hardware problem. It is a data visibility problem. The production lines in most mid-size plants already carry the cameras and PLC signals needed to track availability, performance, and quality in real time. The missing piece is the edge AI layer that turns those passive inputs into an active, shift-by-shift OEE score.
Nagare by Jidoka Technologies adds that layer to your existing infrastructure, without new sensors, without a cloud dependency, and without a six-month implementation project. If your current real time OEE monitoring is still happening in a spreadsheet the morning after the shift, request a Jidoka deployment audit this week and seeing exactly what OEE data your floor is already generating.
Frequently Asked Questions
1. What is real-time OEE monitoring?
Real-time OEE monitoring is the automated tracking of availability, performance, and quality metrics directly from production machines or cameras, updated continuously rather than at shift end. It enables immediate intervention when any OEE component drops below threshold, rather than discovering losses after the fact in a next-day report.
2. Can I monitor OEE without adding new sensors?
Yes. Modern edge AI platforms can extract OEE data from existing cameras and PLC signals already present on most production lines. Camera-based AI detects machine state, cycle time, and defect events without hardware modifications. Systems like Nagare by Jidoka Technologies deploy using existing CCTV infrastructure and can be live within days for a single-line deployment.
3. What is a good OEE score for manufacturing?
World-class manufacturers target 85%+ OEE. The global manufacturing average is approximately 60% (MachineCDN, 2026). A score below 65% typically indicates significant unaddressed availability or performance loss. Most mid-size manufacturers see the fastest gains by closing the micro-stop gap in the performance component first, as it is both the largest unrecognised loss and the easiest to address with camera-based monitoring.
4. How long does it take to deploy OEE monitoring?
Deployment timelines vary significantly. Sensor-based enterprise platforms typically take 3 to 6 months, per MachineCDN 2026 data. Camera-based edge AI systems like Nagare deploy on existing infrastructure without IT network changes or cloud dependency. For most mid-size plants, a single-line deployment can be live within 1 to 2 weeks from the decision to proceed.
